Various aspects of user preference learning and recommender systems
|
|
- Stephanie Marshall
- 6 years ago
- Views:
Transcription
1 Various aspects of user preference learning and recommender systems Alan Eckhardt Department of Software Engineering, Charles University, Institute of Computer Science, Czech Academy of Science, Prague, Czech Republic Abstract. In this paper, we describe area of recommender systems, with focus on user preference learning problem. We describe such system and identify some interesting problems. We will compare how well different approaches cope with some of the problems. This paper may serve as an introduction to the area of user preference learning with a hint on some interesting problems that have not been solved yet. Keywords: user preference learning, data mining, recommender systems 1 Introduction User preference learning is an important part of any recommender system. We will work with a scenario of user searching for some object (we will refer to user as she for not having to distinguish between he and she). Recommendation may help user to find what she is looking for more quickly and efficiently, because she has not to crawl through hundreds of products but sees the recommended products on only one or two pages. Of course, these recommended products may not be an exhaustive list, but they are a hint for user. In Section 2 some important related work is studied, providing also an introduction to the problematic of user modeling. Then, in Section 3, we describe how user preferences are modeled in our approach. In Section 4 is described a typical scenario of recommendation cycle for user. We also describe how our user model is constructed and ways for estimating usefulness of a user model. In Section 5 are listed some interesting problems associated with learning of user preferences. Finally, in Section 6 are conclusive remarks and more importantly areas for future work in this field are proposed. 1.1 Example In the whole paper, we will refer to a set of objects. These objects are supposed to be of interest for user, probably she wants to buy one. In our traditional example, user is This paper was supported by Czech projects MSM and 1ET K. Richta, J. Pokorný, V. Snášel (Eds.): Dateso 2009, pp , ISBN
2 Various aspects of user preference learning and recommender systems 57 buying a notebook. She has some preferences of notebooks, e.g. the maximal price she is willing to pay, the preferred manufacturer or the size of the display. This example is suitable for our approach because notebooks have well defined attributes that describes the product completely. More about this is in Section Notation We will work with a set of objects X. Set X can be also viewed as a set of identifiers of objects (id), which will be often referred to as o. Every object has attributes A 1,...,A N with domains D A1,...,D AN. If we want to specify the value of an attribute A i for an object o j, we will use notation A i (o j ). We will use X i (a) when denoting a set of objects for which attribute A i has the value a. When the attribute will be clear from context (which will be most of the times), we will use only X(a). 2 Related work User preference modeling was very nicely described in [4] and also in a more general view in [10]. In Figure 1 (which was taken from [4]) are various components of preference modeling. Model is how user preferences are understood for our purposes Total order of outcomes (or objects) will be most suitable. That means we can create a list of all objects ordered according to user preferences. Language is a way for user to express her preferences. It may be a rating of an object, as V (o) in Figure 1, or a query to the system etc. Language is explored in Section 4.1. The most interesting part for us is Interpretation, where the information from user is somehow transformed into Model, e.g. the total order of outcomes. However, because of intuition, we will slightly change the notation we will refer to the method for creating the total order as user model or user preference model. Interpretation may be also viewed as learning phase, where a user preference model is constructed. In the following two sections, two alternative ways of user modeling are described along with their possible interpretations. First, qualitative models are based on comparing two objects between them, and second, quantitative models are based on evaluation of a single object with a scoring function. 2.1 Qualitative approaches Preference relations are the most used and studied qualitative approach. There is a huge amount of related work in the field of preference relations. Preference relations represent preferences as a relation between two objects, it is usually assumed that this relation creates some pre-order on X. There are typically three relations, P are strict preferences, I represents indistinguishability of objects or equality of preference and R is union of P and I meaning that it represents non-strict preferences. For example P(o 1,o 2 ) means that o 1 is strictly preferred to o 2, I(o 1,o 2 ) on the other hand means that o 1 and o 2 have the same preference and finally R(o 1,o 2 ) means that o 1 is preferred or equal to o 2.
3 58 Alan Eckhardt Fig. 1. Preference model components. Preference relations in database systems and their integration into SQL by preference queries was studied by Chomicki in [8], [9]. Also Kießling contributed to this field with [25]. A different approach was suggested by Kießling in [19], [24]. This approach is based on the idea of preference relations but it uses relations over attribute values rather than relations over whole objects. This is more like our approach based on fuzzy logic. However Kießling does not use scoring functions but uses special predicates POS, NEG etc. to represent relation between two attribute values. An example from [24] is from the area of cars: POS(transmission, automatic) and NEG(make, Ferrari) meaning that automatic transmission is preferred to any other type and any maker is preferred to Ferrari. As for learning of preference relations, a great contribution is from Fürnkranz and Hüllermeier [17], [20]. 2.2 Quantitative approaches The other approach, also adopted by us, is quantitative. It sorts objects by a score defined by a scoring function. This approach is arguably less expressive than the qualitative one it can not express e.g. a cycle in preferences. There are also some very interesting works in this area. Content based models Content based models uses attributes of object for construction of scoring function. For example Fagin in [16] proposed a way of combining numerous fuzzy inputs. Another classical work is from Agrawal [6]. Collaborative filtering Besides content based models, such as the one presented in Section 3, there is another widely used user model that is based on the preferences of
4 Various aspects of user preference learning and recommender systems 59 other users. Collaborative filtering was proposed in early 90 s in [18] and further developed. One of the well-known systems using collaborative filtering is GroupLens [27]. Collaborative filtering is based on the idea of similarity of users. When we want to know how user u 1 will like object o 1, one way is to look how other people liked o 1. Amazon.com succeeds in describing this approach in one sentence Customers Who Bought This Item Also Bought.... The better way is to restrict only to those users that are similar to u 1. The similarity may be computed in various way, the most common is the similarity of ratings of objects other than o 1. Other possibility is to compute the similarity of user profiles e.g. find managers, from 25 to 30, divorced, with interest in psychology and computer science. There is a hidden assumption that similarity in profile imply similarity in preferences, which may not be always true. 3 User model based on fuzzy logic In this section, we describe user model we are using. This model is based on a scoring function that assigns every object a score that represents the rating of that object. User rating of an object is a fuzzy subset of X, i.e. a function R(o) : X [0, 1], where 0 means least preferred and 1 means most preferred. Our scoring function is divided into two steps. Local preferences In the first step, which we call local preferences, every attribute value of object o is normalized using a fuzzy set f i : D Ai [0,1]. The meaning is that 1 represents most preferred value and 0 stands for the least preferred value. These fuzzy sets are also called objectives or preferences over attributes. With this transformation, N the original space of objects attributes D Ai is transformed into [0, 1] N. Moreover, i=1 we know that the object with transformed attribute values equal to [1,...,1] is the most preferred object. It probably does not exist in the real world, though. On the other side, the object with values [0,...,0] is the least preferred, which is more probable to be found in reality. Global preferences In the second step, called global preferences, the normalized attributes are aggregated into the overall score of the object using an aggregation : [0,1] N [0,1]. Aggregation function is also often called utility function. Aggregation function may take different forms; one of the most common is weighted average, as in the following = (2 f Price (o) + 1 f Display (o) + 3 f HDD (o) + 1 f RAM (o))/7, where f A are fuzzy sets for normalization of attribute A. Another totally different approach was proposed in [15]. It uses the training dataset as partitioning of normalized space [0, 1] N. For example, if we have an object with normalized values [0.4, 0.2, 0.5] with rating 3, every other object with better attribute values (e.g. [0.5,0.4,0.7]) is supposed to have rating at least 3. In this way, we can find the highest lower bound on any object with unknown rating. In [15] was also proposed
5 60 Alan Eckhardt a method for interpolation of ratings between the objects with known ratings and even using the ideal (non-existent) virtual object with normalized values [1,..., 1] with rating 6. In other words, we can say that the pareto front is constructed in the first step. Pareto front is a set of objects that are not dominated by any other object. We say that object o 1 dominates object o 2 iff i = 1,...,N : f i (o 1 ) > f i (o 2 ), i.e. o 1 is better in every attribute than o 2. In the second step, we choose the best object from the pareto front. 4 A recommender system A recommender system tries to help user to find the object she is looking for. It is necessary for user to transfer some information about her preferences to the system. It is convenient for user to describe her preferences in an intuitive and simple way. The more complex user interface is, the more structured information the system gets but much less users will use it (according to [26], as little as 9 out of 260 people provided a feedback to their system). This fact also penalizes preference relations user is supposed to compare two objects, but the number of couples is quadratic to the number of objects. There is an example how a recommender system may work in Figure 2. System presents user with a set of objects S 0. User rates some of these objects and this information is sent back to the system as feedback U 0. User model is constructed from user s ratings and a personalized set S 1 is sent to user. Again, user rates some objects (U 1 ) and system updates its user model and sends S 2 and this cycle may go on until user is satisfied or bored. In Section 4.1 various possible types of feedback from user are studied. In Section 4.3 and 4.2 the construction and update of user model is described. Recommender system User Construction of user model Update of user model S 0 S 1 U 0 U 1 User decision making User decision making S 2... Fig. 2. A use of a recommender system in steps.
6 Various aspects of user preference learning and recommender systems 61 The process of recommendation is in Figure 3. User does some actions with the system, which are processed by various components of the system. You can see that some inputs may be processed by multiple components. There are three examples in the figure analysis of user behaviour, collaborative filtering, analysis of ratings and direct query. Each of these components then creates user model which is used for prediction of preference of all objects. Some models, like collaborative filtering, use the information about other users or other additional information. All these models are then combined together to provide most precise recommendation for user. Furthermore, when the system identifies some of the situation discussed in 5, it can favour the model that behaves best in this situation or the other way round. Collaborative filtering is not good when there is a small number of users, so if that is the case, it can be disfavoured. Recommended system Database with other users preferences Analysis of ratings Collaborative filtering Combination of different methods Analysis of behaviour Direct query Ratings Behaviour Queries Restrictions of attributes User Recommendation of objects Fig. 3. Structure of a recommender system. 4.1 Input from user From the information user provides to the system, her user model is built. User model should be capable to determine which objects user will like or to what degree an object will be preferred. The construction of user model is of most interest for us. We are working with user ratings. These ratings user associates to a small number of objects. This is key aspect of user model construction it can not be expected that user will rate hundreds of objects. When doing experiments, we often limit the size of training set to 40 objects. Other approaches may expect different forms of information from user other than ratings. For example for preference relations, comparisons between two objects is the
7 62 Alan Eckhardt expected input. The full-text query issued by user may be also viewed as a source of information about what user wants document retrieval uses queries as its only information from user. Datasets There exists publicly available datasets of user ratings such as Netflix [3]. In these datasets exist users with even thousands of ratings. Unfortunately, most of these datasets have a very small number of attributes. Books [5] - have author, title, year of publication and publisher. It contains ratings with non-zero rating from users. Jokes [23] - have no attributes, except the text of the joke itself. It contains 4.1 million of ratings by users. Films [2], [3] - have many attributes (from IMDB [1]) but they are complicated. One film can have many actors, many producers, many directors etc. The normal attributes such as length of a film are not determined easily, because they differ across countries, editions or releases. Movielens contains 10 million ratings from users. Netflix contains over 100 million ratings. All this is data from users who were using a system for a long time, several years in most cases. Behaviour analysis User behaviour interpretation was studied in [21] and [22]. Because user tends not to give very much information about herself, the interpretation of her behaviour may provide a useful information that supports the explicit actions she had done (such as ratings of objects). There are many events that can be monitored, such as the time spent on a web page with details of an object, clicking on a page, scrolling down a document, filtering the content of the page, issuing a query etc. These actions are then interpreted as if they were motivated by her preferences, e.g. the longer user stays on a page, the more preferred is the object on that page. When user browses the shop by categories or restricts values of some attribute, it is a clear statement what she likes. For example when user narrows her search to notebooks with display size 14, we can deduce that 14 is the best size of display. This information helps when creating local preferences (in Section 4.2). The order in which such restrictions are applied may represent importance of attributes. Typical counter-example against interpretation of behaviour is when user is going for a coffee, leaving the browser open on that page, or a user searching for an object for a friend (see also Section 5 for this issue). In following sections, we outline some methods for creating local and global preferences, which are contributions we made to this area in the past. 4.2 Learning local preferences The acquisition of local preferences differs for different types of attributes. For nominal attributes, we use a method based on representative value that is computed from user ratings [11, 13]. For numerical attributes, the problem is more complicated there is usually only one object with value a. A typical example is price there is often only one notebook
8 Various aspects of user preference learning and recommender systems 63 with price e.g. 941$. We can stick with traditional methods such as linear regression, where the input is formed from the ratings of objects and their prices. However linear regression may be affected by the distribution of data, but we want a function that corresponds to real values across the whole domain, not only there where are most values. Because of this, we proposed a way of using representants for numerical domain in [11]. Lately, we identified another problem with numerical domains. It may happen that a value of another attribute, often nominal, affects the normalization of a numerical attribute. For example, when flying with British Airways, you may prefer lower price, because the comfort is fine in economy class, but with Aeroflot, you may prefer higher price because the overall comfort is worse. This phenomenon is related to ceteris paribus preferences [28] and CP-nets [7]. 4.3 Learning global preferences Method called Statistical was described in [11] and in [12]. It is based on the evaluation of distribution of ratings across attribute domain D A and from this distribution it is possible to derive weight that A plays in the decision process of user. Then the weighted average with these weights is used as aggregation function. When constructing scoring function called Instances that uses the objects from training set as lower bounds (from [15]), there is little to do. It does not need any further transformation or analysis. This is unfortunately balanced by a more complicated computation during evaluation of new objects. 4.4 How to measure usefulness of a recommender system In this section we identify several ways of measuring usefulness of a recommender system. As in data-mining, we adopt the idea of training (Tr) and testing sets (Ts). User model is constructed from objects in training set and then its performance is measured on testing set. In the following, the real preference of an object o will be denoted as R(o) and preference user model estimates as R(o). RMSE RMSE stands for root mean squared error. It is widely used as error measure in data-mining community. It is computed as o Ts ( R(o) R(o)) 2 / Ts. When considering user preferences, we introduced a modified RMSE, weighted RMSE. It associates more weight to the preferred objects than to the non-preferred. The formula is o Ts R(o) ( R(o) R(o)) 2 / o Ts R(o). The less is RMSE, the better the system performs. Tau coefficient Tau coefficient is known in economy. It is used to measure the similarity of the ordered lists of objects. It is assumed that both lists contain the same set of objects. For our purpose, we compare the ordering of objects by real user preferences R(o) with the ordering user model would make R(o).
9 64 Alan Eckhardt Tau coefficient is based on concordant and discordant pairs. A pair (o 1,o 2 ),(p 1,p 2 ) is concordant, if sgn(r(o 1 ) R(o 2 )) = sgn( R(p 1 ) R(p 2 )), where sgn is signum function. Then the coefficient is computed as τ = nc n d where n 1 2 n(n 1) c is number of concordant pairs, n d is the number of discordant pairs and n is the number of objects in lists. The higher Tau coefficient is, the better for the system. We can apply weighting scheme to Tau coefficient, too. Objects with higher real preferences will matter more than objects with low preference, when comparing the two orderings. ROC curves ROC curve is a method for capturing the performance of a system under different conditions. In terms of classification of positive and negative examples, it tells how the recall (the ratio between the number of correctly classified positive objects and the total number of positive objects) of the system increases if we relax the ratio of correctly classified negative objects and the number of all negative objects. The problem here is that we do not have a simple positive-negative scenario. Ratings have typically the domain {1,2,3,4,5}. The possible solution is to consider several cuts in this domain and measure at each of these cuts. We can take as positive objects with rating 5 and the rest as negative. Then, objects with ratings 5 and 4 will be considered positive and the rest as negative etc. In this way, we will get 5 different ROC curves. Amount of information required from user Another important characteristic of a recommender system is amount of information that is required from user. Possible approaches are described in Section 4.1. Often, the higher precision of a system is outweighed by a larger investment from user. But the intuition says that few users will be willing to perform complicated tasks, no matter the increase of helpfulness of the system. The less information the system needs from user, the better for user. Improvement in user ease of work with the system The main task of recommender systems is to help the user. All the above mentioned measures are good because they can be computed analytically, but they cannot capture the real improvement for user. This improvement can be only given by the real user working with the various recommender systems and comparing them between them. There are some problems with this comparison Every man perceives the ease of work differently. There should be a large number of people testing the systems to be compared to evaluate overall suitability. The system has to be implemented including the user interface. Different methods work with different inputs from the user and they have to be able to monitor these inputs. People testing the system has to have a motivation to work with the system. If they are not, they would not be critical or they would apply some criteria irrelevant in the real usage of the system. This is most difficult to achieve.
10 Various aspects of user preference learning and recommender systems 65 5 Interesting problems of recommender systems In this section, some interesting problems are described. All these problems relate to user preference learning and can be viewed as the main contribution of this paper, besides the introductory part. Cold start When a system is based on user ratings, it has serious troubles in the beginning of its existence. This problem is most pronounced for collaborative filtering-based system, because they need a lot of rating from a lot of users for making correct predictions. For content based recommendation, this is not much of a problem because the accuracy of recommendation does not depend on the number of users of the system. The following issue however affect both approaches. New user When a new user starts using the system, it is hard to recommend something because the system has little information about her. This issue may be overcome with a default user profile, but this solution is not personalized enough. There is a possibility to learn something quickly about a new user it is by choosing a good set S 0 in Figure 2. When there are very different objects in S 0, we can learn immediately what user certainly does not like and what area is of her interest. When S 0 is constructed randomly, there is a higher possibility that there would not be any object the area of user s interest. Suitable S 0 may be found for example using clustering. Identifying context for user When a user works with the system, it is assumed that she searches for some object she wants. But there are other possible motives such as scanning the area or finding some object for a friend. This context of work deeply influences user behaviour and may severely damage existing user model. Small rating scale Typical rating is expressed on the scale {1,2,3,4,5}. When having thousands of objects, this scale is far too small for capturing the ordering among the best objects. We proposed a method Phases for overcoming this obstacle in [14], but it is not a complete analyze and there are many more aspects of this problem. Phases are suited for a single session, for long-term usage there should be another way of differentiating among the objects with rating 5. This problem could be of course solved by enlarging the scale of ratings to e.g. {1,...,100} but its benefit is questionable. Will user really feel the difference between ratings 64 and 65? Will user be able to apply it consistently? Negative preferences In our model, 0 means the lowest preference. When a manufacturer ASUS has the preference 0, it penalizes the notebook, but in the overall score it may be outweighed by other attributes which are more preferred by user. Negative preferences are typically more stronger than this a notebook made by ASUS is strongly disregarded. The strength may be expressed by the weight of the attribute, but it is no solution, because for other manufacturers, this attribute may not have such a big weight. Our task is to model a better way of penalizing objects with a highly non preferred value and the way of expressing such strong negative preference.
11 66 Alan Eckhardt Time dimension Main problem with acquisition of user preferences is the small size of training data. When user is using the system for a longer period, the training data may get bigger. But here comes a new problem user preferences are typically unstable. What user preferred a month ago may not be good today. But the high rating user provided a month ago is the same high rating provided today. There is a need to find a method for penalization of older ratings or promotion of newer ones. The system needs to guess whether to apply the penalization at all maybe the preferences stay the same when buying a house, but they may change quickly for a movie. 6 Conclusion We have described a recommender system and identified several interesting issues and problems that occur during user model construction. The main contribution is the suggestion of several possible ways how to measure helpfulness of a recommender system. The experimental evaluation is often lacking in more theoretical focused papers. Also the list of some problems or typical situations in recommender systems may be inspiring for searching of their solutions. We hope that this analysis would be helpful to anyone new in recommender systems and user modeling field. 6.1 Issues for future work All problems in Section 5 are worth studying, but we would like to address the two last problems in near future that of Negative preferences and Time dimension. References 1. The internet movie database Movielens dataset Netflix dataset Preference handling an introductory tutorial. brafman/tutorial.pdf. 5. Improving Recommendation Lists Through Topic Diversification, R. Agrawal and E. L. Wimmers. A framework for expressing and combining preferences. SIGMOD Rec., 29(2): , C. Boutilier, R. I. Brafman, H. H. Hoos, and D. Poole. Cp-nets: A tool for representing and reasoning with conditional ceteris paribus preference statements. Journal of Artificial Intelligence Research, 21:2004, J. Chomicki. Preference queries. CoRR, cs.db/ , J. Chomicki. Preference formulas in relational queries. ACM Trans. Database Syst., 28(4): , J. Doyle. Prospects for preferences. Computational Intelligence, 20: (26), May A. Eckhardt. Inductive models of user preferences for semantic web. In J. Pokorný, V. Snášel, and K. Richta, editors, DATESO 2007, volume 235 of CEUR Workshop Proceedings, pages Matfyz Press, Praha, 2007.
12 Various aspects of user preference learning and recommender systems A. Eckhardt, T. Horváth, D. Maruščák, R. Novotný, and P. Vojtáš. Uncertainty issues in automating process connecting web and user. In P. C. G. da Costa, editor, URSW 07 Uncertainty Reasoning for the Semantic Web - Volume 3, pages The 6th International Semantic Web Conference, A. Eckhardt, T. Horváth, and P. Vojtáš. Learning different user profile annotated rules for fuzzy preference top-k querying. In H. Prade and V. Subrahmanian, editors, International Conference on Scalable Uncertainty Management, volume 4772 of Lecture Notes In Computer Science, pages , Washington DC, USA, Springer Berlin / Heidelberg. 14. A. Eckhardt, T. Horváth, and P. Vojtáš. PHASES: A user profile learning approach for web search. In T. Lin, L. Haas, R. Motwani, A. Broder, and H. Ho, editors, 2007 IEEE/WIC/ACM International Conference on Web Intelligence - WI 2007, pages IEEE, A. Eckhardt and P. Vojtáš. Considering data-mining techniques in user preference learning. In 2008 International Workshop on Web Information Retrieval Support Systems, R. Fagin. Combining fuzzy information from multiple systems. pages , J. Fürnkranz and E. Hüllermeier. Pairwise preference learning and ranking. In In Proc. ECML-03, Cavtat-Dubrovnik, pages Springer-Verlag, D. Goldberg, D. Nichols, B. M. Oki, and D. Terry. Using collaborative filtering to weave an information tapestry. Commun. ACM, 35(12):61 70, S. Holland, M. Ester, and W. Kiessling. Preference mining: A novel approach on mining user preferences for personalized applications. In Knowledge Discovery in Databases: PKDD 2003, pages Springer Berlin / Heidelberg, E. Hüllermeier and J. Fürnkranz. Learning preference models from data: On the problem of label ranking and its variants. In G. D. Riccia, D. Dubois, R. Kruse, and H. Lenz, editors, Preferences and Similarities, pages Springer-Verlag, T. Joachims. Optimizing search engines using clickthrough data. In KDD 02: Proceedings of the eighth ACM SIGKDD international conference on Knowledge discovery and data mining, pages , New York, NY, USA, ACM Press. 22. T. Joachims, L. Granka, B. Pan, H. Hembrooke, and G. Gay. Accurately interpreting clickthrough data as implicit feedback. In SIGIR 05: Proceedings of the 28th annual international ACM SIGIR conference on Research and development in information retrieval, pages , New York, NY, USA, ACM. 23. G. K. Eigentaste: A constant time collaborative filtering algorithm. Information Retrieval, 4: (19), July W. Kiessling. Foundations of preferences in database systems. In VLDB 02: Proceedings of the 28th international conference on Very Large Data Bases, pages VLDB Endowment, W. Kiessling and G. Köstler. Preference sql: design, implementation, experiences. In VLDB 02: Proceedings of the 28th international conference on Very Large Data Bases, pages VLDB Endowment, S. E. Middleton, N. Shadbolt, and D. D. Roure. Capturing interest through inference and visualization: Ontological user profiling in recommender systems. In K-CAP2003, A. M. Rashid, I. Albert, D. Cosley, S. K. Lam, S. M. McNee, J. A. Konstan, and J. Riedl. Getting to know you: learning new user preferences in recommender systems. In IUI 02: Proceedings of the 7th international conference on Intelligent user interfaces, pages , New York, NY, USA, ACM. 28. G. H. Wright. The logic of preference reconsidered. In Theory and Decision, volume 3, pages , 1972.
PrefWork - a framework for the user preference learning methods testing
PrefWork - a framework for the user preference learning methods testing Alan Eckhardt 1,2 1 Department of Software Engineering, Charles University, 2 Institute of Computer Science, Czech Academy of Science,
More informationTransformation and aggregation preprocessing for top-k recommendation GAP rules induction
Transformation and aggregation preprocessing for top-k recommendation GAP rules induction Marta Vomlelova, Michal Kopecky, Peter Vojtas Faculty of Mathematics and Physics Charles University in Prague Malostranske
More informationCombination of TA- and MD-algorithm for Efficient Solving of Top-K Problem according to User s Preferences
Combination of TA- and MD-algorithm for Efficient Solving of Top-K Problem according to User s Preferences Matúš Ondreička and Jaroslav Pokorný Department of Software Engineering, Faculty of Mathematics
More informationKripke style Dynamic model for Web Annotation with Similarity and Reliability
Kripke style Dynamic model for Web Annotation with Similarity and Reliability M. Kopecký 1, M. Vomlelová 2, P. Vojtáš 1 Faculty of Mathematics and Physics Charles University Malostranske namesti 25, Prague,
More informationIJREAT International Journal of Research in Engineering & Advanced Technology, Volume 1, Issue 5, Oct-Nov, ISSN:
IJREAT International Journal of Research in Engineering & Advanced Technology, Volume 1, Issue 5, Oct-Nov, 20131 Improve Search Engine Relevance with Filter session Addlin Shinney R 1, Saravana Kumar T
More informationUse of KNN for the Netflix Prize Ted Hong, Dimitris Tsamis Stanford University
Use of KNN for the Netflix Prize Ted Hong, Dimitris Tsamis Stanford University {tedhong, dtsamis}@stanford.edu Abstract This paper analyzes the performance of various KNNs techniques as applied to the
More informationEstimating Credibility of User Clicks with Mouse Movement and Eye-tracking Information
Estimating Credibility of User Clicks with Mouse Movement and Eye-tracking Information Jiaxin Mao, Yiqun Liu, Min Zhang, Shaoping Ma Department of Computer Science and Technology, Tsinghua University Background
More informationEfficient Case Based Feature Construction
Efficient Case Based Feature Construction Ingo Mierswa and Michael Wurst Artificial Intelligence Unit,Department of Computer Science, University of Dortmund, Germany {mierswa, wurst}@ls8.cs.uni-dortmund.de
More informationJoint Entity Resolution
Joint Entity Resolution Steven Euijong Whang, Hector Garcia-Molina Computer Science Department, Stanford University 353 Serra Mall, Stanford, CA 94305, USA {swhang, hector}@cs.stanford.edu No Institute
More informationAn Intelligent Clustering Algorithm for High Dimensional and Highly Overlapped Photo-Thermal Infrared Imaging Data
An Intelligent Clustering Algorithm for High Dimensional and Highly Overlapped Photo-Thermal Infrared Imaging Data Nian Zhang and Lara Thompson Department of Electrical and Computer Engineering, University
More informationMining High Order Decision Rules
Mining High Order Decision Rules Y.Y. Yao Department of Computer Science, University of Regina Regina, Saskatchewan, Canada S4S 0A2 e-mail: yyao@cs.uregina.ca Abstract. We introduce the notion of high
More informationSingular Value Decomposition, and Application to Recommender Systems
Singular Value Decomposition, and Application to Recommender Systems CSE 6363 Machine Learning Vassilis Athitsos Computer Science and Engineering Department University of Texas at Arlington 1 Recommendation
More informationFoundations of Preferences in Database Systems
Foundations of Preferences in Database Systems Werner Kießling Institute of Computer Science University of Augsburg W. Kießling, 28th Intern. Conf. on Very Large Databases (VLDB 2002), Hong Kong, 21.08.02
More informationA Constrained Spreading Activation Approach to Collaborative Filtering
A Constrained Spreading Activation Approach to Collaborative Filtering Josephine Griffith 1, Colm O Riordan 1, and Humphrey Sorensen 2 1 Dept. of Information Technology, National University of Ireland,
More informationCP nets: representing and reasoning with preferences of multiple agents
CP nets: representing and reasoning with preferences of multiple agents F. Rossi and K. B. Venable Dept. of Mathematics University of Padova Padova Italy frossikvenable @math.unipd.it Abstract We introduce
More informationOptimizing Search Engines using Click-through Data
Optimizing Search Engines using Click-through Data By Sameep - 100050003 Rahee - 100050028 Anil - 100050082 1 Overview Web Search Engines : Creating a good information retrieval system Previous Approaches
More informationNext Stop Recommender
Next Stop Recommender Ben Ripley, Dirksen Liu, Maiga Chang, and Kinshuk School of Computing and Information Systems Athabasca University Athabasca, Canada maiga@ms2.hinet.net, kinshuk@athabascau.ca Abstract
More informationProwess Improvement of Accuracy for Moving Rating Recommendation System
2017 IJSRST Volume 3 Issue 1 Print ISSN: 2395-6011 Online ISSN: 2395-602X Themed Section: Scienceand Technology Prowess Improvement of Accuracy for Moving Rating Recommendation System P. Damodharan *1,
More informationSemantic Preference Retrieval for Querying Knowledge Bases
Semantic Preference Retrieval for Querying Knowledge Bases Christian Scheel scheel@dai-lab.de Alan Said alan@dai-lab.de DAI-Labor Technische Universität Berlin, D-10587, Germany Sahin Albayrak sahin@dai-lab.de
More informationComparison of Recommender System Algorithms focusing on the New-Item and User-Bias Problem
Comparison of Recommender System Algorithms focusing on the New-Item and User-Bias Problem Stefan Hauger 1, Karen H. L. Tso 2, and Lars Schmidt-Thieme 2 1 Department of Computer Science, University of
More informationModel theoretic and fixpoint semantics for preference queries over imperfect data
Model theoretic and fixpoint semantics for preference queries over imperfect data Peter Vojtáš Charles University and Czech Academy of Science, Prague Peter.Vojtas@mff.cuni.cz Abstract. We present an overview
More informationRanking Web Pages by Associating Keywords with Locations
Ranking Web Pages by Associating Keywords with Locations Peiquan Jin, Xiaoxiang Zhang, Qingqing Zhang, Sheng Lin, and Lihua Yue University of Science and Technology of China, 230027, Hefei, China jpq@ustc.edu.cn
More informationQuery-Sensitive Similarity Measure for Content-Based Image Retrieval
Query-Sensitive Similarity Measure for Content-Based Image Retrieval Zhi-Hua Zhou Hong-Bin Dai National Laboratory for Novel Software Technology Nanjing University, Nanjing 2193, China {zhouzh, daihb}@lamda.nju.edu.cn
More informationRank Measures for Ordering
Rank Measures for Ordering Jin Huang and Charles X. Ling Department of Computer Science The University of Western Ontario London, Ontario, Canada N6A 5B7 email: fjhuang33, clingg@csd.uwo.ca Abstract. Many
More informationA PERSONALIZED RECOMMENDER SYSTEM FOR TELECOM PRODUCTS AND SERVICES
A PERSONALIZED RECOMMENDER SYSTEM FOR TELECOM PRODUCTS AND SERVICES Zui Zhang, Kun Liu, William Wang, Tai Zhang and Jie Lu Decision Systems & e-service Intelligence Lab, Centre for Quantum Computation
More informationAdvances in Natural and Applied Sciences. Information Retrieval Using Collaborative Filtering and Item Based Recommendation
AENSI Journals Advances in Natural and Applied Sciences ISSN:1995-0772 EISSN: 1998-1090 Journal home page: www.aensiweb.com/anas Information Retrieval Using Collaborative Filtering and Item Based Recommendation
More informationBuilding an Infrastructure for Law Enforcement Information Sharing and Collaboration: Design Issues and Challenges
Submission to the National Conference on Digital Government, 2001 Submission Type: PAPER (SHORT PAPER preferred) Building an Infrastructure for Law Enforcement Information Sharing and Collaboration: Design
More informationYunfeng Zhang 1, Huan Wang 2, Jie Zhu 1 1 Computer Science & Engineering Department, North China Institute of Aerospace
[Type text] [Type text] [Type text] ISSN : 0974-7435 Volume 10 Issue 20 BioTechnology 2014 An Indian Journal FULL PAPER BTAIJ, 10(20), 2014 [12526-12531] Exploration on the data mining system construction
More informationCADIAL Search Engine at INEX
CADIAL Search Engine at INEX Jure Mijić 1, Marie-Francine Moens 2, and Bojana Dalbelo Bašić 1 1 Faculty of Electrical Engineering and Computing, University of Zagreb, Unska 3, 10000 Zagreb, Croatia {jure.mijic,bojana.dalbelo}@fer.hr
More informationTowards a hybrid approach to Netflix Challenge
Towards a hybrid approach to Netflix Challenge Abhishek Gupta, Abhijeet Mohapatra, Tejaswi Tenneti March 12, 2009 1 Introduction Today Recommendation systems [3] have become indispensible because of the
More informationA Chosen-Plaintext Linear Attack on DES
A Chosen-Plaintext Linear Attack on DES Lars R. Knudsen and John Erik Mathiassen Department of Informatics, University of Bergen, N-5020 Bergen, Norway {lars.knudsen,johnm}@ii.uib.no Abstract. In this
More informationA Constrained Spreading Activation Approach to Collaborative Filtering
A Constrained Spreading Activation Approach to Collaborative Filtering Josephine Griffith 1, Colm O Riordan 1, and Humphrey Sorensen 2 1 Dept. of Information Technology, National University of Ireland,
More informationA Tagging Approach to Ontology Mapping
A Tagging Approach to Ontology Mapping Colm Conroy 1, Declan O'Sullivan 1, Dave Lewis 1 1 Knowledge and Data Engineering Group, Trinity College Dublin {coconroy,declan.osullivan,dave.lewis}@cs.tcd.ie Abstract.
More informationLetter Pair Similarity Classification and URL Ranking Based on Feedback Approach
Letter Pair Similarity Classification and URL Ranking Based on Feedback Approach P.T.Shijili 1 P.G Student, Department of CSE, Dr.Nallini Institute of Engineering & Technology, Dharapuram, Tamilnadu, India
More informationSemi-Supervised Clustering with Partial Background Information
Semi-Supervised Clustering with Partial Background Information Jing Gao Pang-Ning Tan Haibin Cheng Abstract Incorporating background knowledge into unsupervised clustering algorithms has been the subject
More informationCombining Review Text Content and Reviewer-Item Rating Matrix to Predict Review Rating
Combining Review Text Content and Reviewer-Item Rating Matrix to Predict Review Rating Dipak J Kakade, Nilesh P Sable Department of Computer Engineering, JSPM S Imperial College of Engg. And Research,
More informationAI Dining Suggestion App. CS 297 Report Bao Pham ( ) Advisor: Dr. Chris Pollett
AI Dining Suggestion App CS 297 Report Bao Pham (009621001) Advisor: Dr. Chris Pollett Abstract Trying to decide what to eat can be challenging and time-consuming. Google or Yelp are two popular search
More informationRecommender Systems. Collaborative Filtering & Content-Based Recommending
Recommender Systems Collaborative Filtering & Content-Based Recommending 1 Recommender Systems Systems for recommending items (e.g. books, movies, CD s, web pages, newsgroup messages) to users based on
More informationWEB PAGE RE-RANKING TECHNIQUE IN SEARCH ENGINE
WEB PAGE RE-RANKING TECHNIQUE IN SEARCH ENGINE Ms.S.Muthukakshmi 1, R. Surya 2, M. Umira Taj 3 Assistant Professor, Department of Information Technology, Sri Krishna College of Technology, Kovaipudur,
More informationPerformance Measures for Multi-Graded Relevance
Performance Measures for Multi-Graded Relevance Christian Scheel, Andreas Lommatzsch, and Sahin Albayrak Technische Universität Berlin, DAI-Labor, Germany {christian.scheel,andreas.lommatzsch,sahin.albayrak}@dai-labor.de
More informationRecommender Systems 6CCS3WSN-7CCSMWAL
Recommender Systems 6CCS3WSN-7CCSMWAL http://insidebigdata.com/wp-content/uploads/2014/06/humorrecommender.jpg Some basic methods of recommendation Recommend popular items Collaborative Filtering Item-to-Item:
More informationProbabilistic Abstraction Lattices: A Computationally Efficient Model for Conditional Probability Estimation
Probabilistic Abstraction Lattices: A Computationally Efficient Model for Conditional Probability Estimation Daniel Lowd January 14, 2004 1 Introduction Probabilistic models have shown increasing popularity
More informationA User Preference Based Search Engine
A User Preference Based Search Engine 1 Dondeti Swedhan, 2 L.N.B. Srinivas 1 M-Tech, 2 M-Tech 1 Department of Information Technology, 1 SRM University Kattankulathur, Chennai, India Abstract - In this
More informationDecision Making Procedure: Applications of IBM SPSS Cluster Analysis and Decision Tree
World Applied Sciences Journal 21 (8): 1207-1212, 2013 ISSN 1818-4952 IDOSI Publications, 2013 DOI: 10.5829/idosi.wasj.2013.21.8.2913 Decision Making Procedure: Applications of IBM SPSS Cluster Analysis
More informationQuery Disambiguation from Web Search Logs
Vol.133 (Information Technology and Computer Science 2016), pp.90-94 http://dx.doi.org/10.14257/astl.2016. Query Disambiguation from Web Search Logs Christian Højgaard 1, Joachim Sejr 2, and Yun-Gyung
More informationA New Technique to Optimize User s Browsing Session using Data Mining
Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 4, Issue. 3, March 2015,
More informationSentiment analysis under temporal shift
Sentiment analysis under temporal shift Jan Lukes and Anders Søgaard Dpt. of Computer Science University of Copenhagen Copenhagen, Denmark smx262@alumni.ku.dk Abstract Sentiment analysis models often rely
More informationComparison of of parallel and random approach to
Comparison of of parallel and random approach to acandidate candidatelist listininthe themultifeature multifeaturequerying Peter Gurský Peter Gurský Institute of Computer Science, Faculty of Science Institute
More informationInternational Journal of Advanced Research in. Computer Science and Engineering
Volume 4, Issue 1, January 2014 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Design and Architecture
More informationCorrelation Based Feature Selection with Irrelevant Feature Removal
Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 3, Issue. 4, April 2014,
More informationMichele Gorgoglione Politecnico di Bari Viale Japigia, Bari (Italy)
Does the recommendation task affect a CARS performance? Umberto Panniello Politecnico di Bari Viale Japigia, 82 726 Bari (Italy) +3985962765 m.gorgoglione@poliba.it Michele Gorgoglione Politecnico di Bari
More information/633 Introduction to Algorithms Lecturer: Michael Dinitz Topic: Approximation algorithms Date: 11/27/18
601.433/633 Introduction to Algorithms Lecturer: Michael Dinitz Topic: Approximation algorithms Date: 11/27/18 22.1 Introduction We spent the last two lectures proving that for certain problems, we can
More informationFrom Passages into Elements in XML Retrieval
From Passages into Elements in XML Retrieval Kelly Y. Itakura David R. Cheriton School of Computer Science, University of Waterloo 200 Univ. Ave. W. Waterloo, ON, Canada yitakura@cs.uwaterloo.ca Charles
More informationHybrid Feature Selection for Modeling Intrusion Detection Systems
Hybrid Feature Selection for Modeling Intrusion Detection Systems Srilatha Chebrolu, Ajith Abraham and Johnson P Thomas Department of Computer Science, Oklahoma State University, USA ajith.abraham@ieee.org,
More informationEnhancing Cluster Quality by Using User Browsing Time
Enhancing Cluster Quality by Using User Browsing Time Rehab Duwairi Dept. of Computer Information Systems Jordan Univ. of Sc. and Technology Irbid, Jordan rehab@just.edu.jo Khaleifah Al.jada' Dept. of
More informationAutomatic Query Type Identification Based on Click Through Information
Automatic Query Type Identification Based on Click Through Information Yiqun Liu 1,MinZhang 1,LiyunRu 2, and Shaoping Ma 1 1 State Key Lab of Intelligent Tech. & Sys., Tsinghua University, Beijing, China
More informationData integration for customer preference learning
M. Kopecký 1, M. Vomlelová 2, P. Vojtáš 1 Faculty of Mathematics and Physics Charles University Malostranske namesti 25, Prague, Czech Republic 1 {kopecky vojtas}@ksi.mff.cuni.cz 2 marta@ktiml.mff.cuni.cz
More informationIntroduction. Chapter Background Recommender systems Collaborative based filtering
ii Abstract Recommender systems are used extensively today in many areas to help users and consumers with making decisions. Amazon recommends books based on what you have previously viewed and purchased,
More informationA Survey on Various Techniques of Recommendation System in Web Mining
A Survey on Various Techniques of Recommendation System in Web Mining 1 Yagnesh G. patel, 2 Vishal P.Patel 1 Department of computer engineering 1 S.P.C.E, Visnagar, India Abstract - Today internet has
More informationA PROPOSED HYBRID BOOK RECOMMENDER SYSTEM
A PROPOSED HYBRID BOOK RECOMMENDER SYSTEM SUHAS PATIL [M.Tech Scholar, Department Of Computer Science &Engineering, RKDF IST, Bhopal, RGPV University, India] Dr.Varsha Namdeo [Assistant Professor, Department
More informationEfficient Voting Prediction for Pairwise Multilabel Classification
Efficient Voting Prediction for Pairwise Multilabel Classification Eneldo Loza Mencía, Sang-Hyeun Park and Johannes Fürnkranz TU-Darmstadt - Knowledge Engineering Group Hochschulstr. 10 - Darmstadt - Germany
More informationContent-Based Recommendation for Web Personalization
Content-Based Recommendation for Web Personalization R.Kousalya 1, K.Saranya 2, Dr.V.Saravanan 3 1 PhD Scholar, Manonmaniam Sundaranar University,Tirunelveli HOD,Department of Computer Applications, Dr.NGP
More informationHybrid Recommender Systems for Electronic Commerce
From: AAAI Technical Report WS-00-04. Compilation copyright 2000, AAAI (www.aaai.org). All rights reserved. Hybrid Recommender Systems for Electronic Commerce Thomas Tran and Robin Cohen Dept. of Computer
More informationAuthor(s): Rahul Sami, 2009
Author(s): Rahul Sami, 2009 License: Unless otherwise noted, this material is made available under the terms of the Creative Commons Attribution Noncommercial Share Alike 3.0 License: http://creativecommons.org/licenses/by-nc-sa/3.0/
More informationA Language Independent Author Verifier Using Fuzzy C-Means Clustering
A Language Independent Author Verifier Using Fuzzy C-Means Clustering Notebook for PAN at CLEF 2014 Pashutan Modaresi 1,2 and Philipp Gross 1 1 pressrelations GmbH, Düsseldorf, Germany {pashutan.modaresi,
More informationWeighted Alternating Least Squares (WALS) for Movie Recommendations) Drew Hodun SCPD. Abstract
Weighted Alternating Least Squares (WALS) for Movie Recommendations) Drew Hodun SCPD Abstract There are two common main approaches to ML recommender systems, feedback-based systems and content-based systems.
More informationCS 6740: Advanced Language Technologies April 2, Lecturer: Lillian Lee Scribes: Navin Sivakumar, Lakshmi Ganesh, Taiyang Chen.
CS 6740: Advanced Language Technologies April 2, 2010 Lecture 15: Implicit Relevance Feedback & Clickthrough Data Lecturer: Lillian Lee Scribes: Navin Sivakumar, Lakshmi Ganesh, Taiyang Chen Abstract Explicit
More informationEFFICIENT ATTRIBUTE REDUCTION ALGORITHM
EFFICIENT ATTRIBUTE REDUCTION ALGORITHM Zhongzhi Shi, Shaohui Liu, Zheng Zheng Institute Of Computing Technology,Chinese Academy of Sciences, Beijing, China Abstract: Key words: Efficiency of algorithms
More informationIMAGE RETRIEVAL SYSTEM: BASED ON USER REQUIREMENT AND INFERRING ANALYSIS TROUGH FEEDBACK
IMAGE RETRIEVAL SYSTEM: BASED ON USER REQUIREMENT AND INFERRING ANALYSIS TROUGH FEEDBACK 1 Mount Steffi Varish.C, 2 Guru Rama SenthilVel Abstract - Image Mining is a recent trended approach enveloped in
More informationLink Prediction for Social Network
Link Prediction for Social Network Ning Lin Computer Science and Engineering University of California, San Diego Email: nil016@eng.ucsd.edu Abstract Friendship recommendation has become an important issue
More informationDesign and Realization of Data Mining System based on Web HE Defu1, a
4th International Conference on Machinery, Materials and Computing Technology (ICMMCT 2016) Design and Realization of Data Mining System based on Web HE Defu1, a 1 Department of Quartermaster, Wuhan Economics
More informationInternational Journal of Computer Engineering and Applications, Volume IX, Issue X, Oct. 15 ISSN
DIVERSIFIED DATASET EXPLORATION BASED ON USEFULNESS SCORE Geetanjali Mohite 1, Prof. Gauri Rao 2 1 Student, Department of Computer Engineering, B.V.D.U.C.O.E, Pune, Maharashtra, India 2 Associate Professor,
More informationXML Data in (Object-) Relational Databases
XML Data in (Object-) Relational Databases RNDr. Irena Mlýnková irena.mlynkova@mff.cuni.cz Charles University Faculty of Mathematics and Physics Department of Software Engineering Prague, Czech Republic
More informationA Comparative Study of Selected Classification Algorithms of Data Mining
Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 4, Issue. 6, June 2015, pg.220
More informationFlexible-Hybrid Sequential Floating Search in Statistical Feature Selection
Flexible-Hybrid Sequential Floating Search in Statistical Feature Selection Petr Somol 1,2, Jana Novovičová 1,2, and Pavel Pudil 2,1 1 Dept. of Pattern Recognition, Institute of Information Theory and
More informationNew user profile learning for extremely sparse data sets
New user profile learning for extremely sparse data sets Tomasz Hoffmann, Tadeusz Janasiewicz, and Andrzej Szwabe Institute of Control and Information Engineering, Poznan University of Technology, pl.
More informationMinimal Equation Sets for Output Computation in Object-Oriented Models
Minimal Equation Sets for Output Computation in Object-Oriented Models Vincenzo Manzoni Francesco Casella Dipartimento di Elettronica e Informazione, Politecnico di Milano Piazza Leonardo da Vinci 3, 033
More informationContext based Re-ranking of Web Documents (CReWD)
Context based Re-ranking of Web Documents (CReWD) Arijit Banerjee, Jagadish Venkatraman Graduate Students, Department of Computer Science, Stanford University arijitb@stanford.edu, jagadish@stanford.edu}
More informationCS246: Mining Massive Datasets Jure Leskovec, Stanford University
CS246: Mining Massive Datasets Jure Leskovec, Stanford University http://cs246.stanford.edu [Kumar et al. 99] 2/13/2013 Jure Leskovec, Stanford CS246: Mining Massive Datasets, http://cs246.stanford.edu
More information10-701/15-781, Fall 2006, Final
-7/-78, Fall 6, Final Dec, :pm-8:pm There are 9 questions in this exam ( pages including this cover sheet). If you need more room to work out your answer to a question, use the back of the page and clearly
More informationImproving the Efficiency of Fast Using Semantic Similarity Algorithm
International Journal of Scientific and Research Publications, Volume 4, Issue 1, January 2014 1 Improving the Efficiency of Fast Using Semantic Similarity Algorithm D.KARTHIKA 1, S. DIVAKAR 2 Final year
More informationInverted Index for Fast Nearest Neighbour
Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology ISSN 2320 088X IMPACT FACTOR: 5.258 IJCSMC,
More informationFeature-weighted k-nearest Neighbor Classifier
Proceedings of the 27 IEEE Symposium on Foundations of Computational Intelligence (FOCI 27) Feature-weighted k-nearest Neighbor Classifier Diego P. Vivencio vivencio@comp.uf scar.br Estevam R. Hruschka
More informationLearning to Match. Jun Xu, Zhengdong Lu, Tianqi Chen, Hang Li
Learning to Match Jun Xu, Zhengdong Lu, Tianqi Chen, Hang Li 1. Introduction The main tasks in many applications can be formalized as matching between heterogeneous objects, including search, recommendation,
More informationAdvanced Search Techniques for Large Scale Data Analytics Pavel Zezula and Jan Sedmidubsky Masaryk University
Advanced Search Techniques for Large Scale Data Analytics Pavel Zezula and Jan Sedmidubsky Masaryk University http://disa.fi.muni.cz The Cranfield Paradigm Retrieval Performance Evaluation Evaluation Using
More informationFast Associative Memory
Fast Associative Memory Ricardo Miguel Matos Vieira Instituto Superior Técnico ricardo.vieira@tagus.ist.utl.pt ABSTRACT The associative memory concept presents important advantages over the more common
More informationAn Empirical Comparison of Collaborative Filtering Approaches on Netflix Data
An Empirical Comparison of Collaborative Filtering Approaches on Netflix Data Nicola Barbieri, Massimo Guarascio, Ettore Ritacco ICAR-CNR Via Pietro Bucci 41/c, Rende, Italy {barbieri,guarascio,ritacco}@icar.cnr.it
More informationKeywords Data alignment, Data annotation, Web database, Search Result Record
Volume 5, Issue 8, August 2015 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Annotating Web
More informationDevelopment of Prediction Model for Linked Data based on the Decision Tree for Track A, Task A1
Development of Prediction Model for Linked Data based on the Decision Tree for Track A, Task A1 Dongkyu Jeon and Wooju Kim Dept. of Information and Industrial Engineering, Yonsei University, Seoul, Korea
More informationHybrid Recommendation System Using Clustering and Collaborative Filtering
Hybrid Recommendation System Using Clustering and Collaborative Filtering Roshni Padate Assistant Professor roshni@frcrce.ac.in Priyanka Bane B.E. Student priyankabane56@gmail.com Jayesh Kudase B.E. Student
More informationJeff Howbert Introduction to Machine Learning Winter
Collaborative Filtering Nearest es Neighbor Approach Jeff Howbert Introduction to Machine Learning Winter 2012 1 Bad news Netflix Prize data no longer available to public. Just after contest t ended d
More informationAssignment 5: Collaborative Filtering
Assignment 5: Collaborative Filtering Arash Vahdat Fall 2015 Readings You are highly recommended to check the following readings before/while doing this assignment: Slope One Algorithm: https://en.wikipedia.org/wiki/slope_one.
More informationExtension Study on Item-Based P-Tree Collaborative Filtering Algorithm for Netflix Prize
Extension Study on Item-Based P-Tree Collaborative Filtering Algorithm for Netflix Prize Tingda Lu, Yan Wang, William Perrizo, Amal Perera, Gregory Wettstein Computer Science Department North Dakota State
More informationEFFECTIVE EFFICIENT BOOLEAN RETRIEVAL
EFFECTIVE EFFICIENT BOOLEAN RETRIEVAL J Naveen Kumar 1, Dr. M. Janga Reddy 2 1 jnaveenkumar6@gmail.com, 2 pricipalcmrit@gmail.com 1 M.Tech Student, Department of Computer Science, CMR Institute of Technology,
More informationModeling Contextual Factors of Click Rates
Modeling Contextual Factors of Click Rates Hila Becker Columbia University 500 W. 120th Street New York, NY 10027 Christopher Meek Microsoft Research 1 Microsoft Way Redmond, WA 98052 David Maxwell Chickering
More informationEfficient integration of data mining techniques in DBMSs
Efficient integration of data mining techniques in DBMSs Fadila Bentayeb Jérôme Darmont Cédric Udréa ERIC, University of Lyon 2 5 avenue Pierre Mendès-France 69676 Bron Cedex, FRANCE {bentayeb jdarmont
More informationSupport Vector Regression for Software Reliability Growth Modeling and Prediction
Support Vector Regression for Software Reliability Growth Modeling and Prediction 925 Fei Xing 1 and Ping Guo 2 1 Department of Computer Science Beijing Normal University, Beijing 100875, China xsoar@163.com
More informationInference in Hierarchical Multidimensional Space
Proc. International Conference on Data Technologies and Applications (DATA 2012), Rome, Italy, 25-27 July 2012, 70-76 Related papers: http://conceptoriented.org/ Inference in Hierarchical Multidimensional
More informationMining Gradual Dependencies based on Fuzzy Rank Correlation
Mining Gradual Dependencies based on Fuzzy Rank Correlation Hyung-Won Koh and Eyke Hüllermeier 1 Abstract We propose a novel framework and an algorithm for mining gradual dependencies between attributes
More informationPredicting User Ratings Using Status Models on Amazon.com
Predicting User Ratings Using Status Models on Amazon.com Borui Wang Stanford University borui@stanford.edu Guan (Bell) Wang Stanford University guanw@stanford.edu Group 19 Zhemin Li Stanford University
More information